Current surveillance systems operate in a highly dynamic environment in which large numbers of sensors on board
multiple platforms must cooperate in order to achieve overall mission success. In an attempt to maximize sensor
performance, today's sensors employ rudimentary or, in some cases, inflexible sensor tasking schemes. These
approaches are highly tuned to a specific scenario and geometry and are inflexible to changes in the mission,
environmental conditions, heterogeneous sensors, and different system architectures. As the complexity of the problem
space increases and new sensors become available, it is critical to have a sensor management scheme that is capable of
incorporating new environmental knowledge, new sensors and different systems approaches with minimal computational
impact on the overall system. Each system should develop an autonomous sensor tasking capability which factors in
global concerns within the complete distributed network of platforms and sensors. Moreover, tasking efficiency can be
improved by a highly developed understanding of sensor performance at each point in time. This can be achieved by
incorporating the impact of problem geometry - sensor location, track object type and view angle - and weather
phenomena, such as clouds, aerosols, turbulence and sun glint.
This paper describes our approach for simultaneously optimizing sensor resource management, surveillance objectives,
and atmospheric transmission of signals while minimizing sensor and environmental noise. Our approach uses a genetic
algorithm to evolve a population of sensor tasking assignments through constantly-updating track locations, weather
conditions, and lighting conditions. Preliminary studies demonstrate encouraging improvements in sensor management
performance. We will present results from our preliminary studies and discuss a path forward for our technology.
|